Swing It Open, Swing It Shut
A Type 2 decision is like walking through a two-way swinging door at a restaurant. You push through, look around, and if you don't like what's on the other side, you push back through. You're exactly where you started, having lost only the few seconds it took to look.
In engineering, most decisions you make daily are Type 2. The defining characteristic is that the cost of reversal is low relative to the cost of delay. Spending three days debating a choice you could undo in three minutes is a net loss for the organization, even if the debate leads to a marginally "better" answer. You learn more in 30 minutes of trying than in 3 hours of debating.
Characteristics of Type 2 Decisions:
| Property | Description |
|---|---|
| Reversal Cost | Low. A git revert, a config change, a feature flag toggle. |
| Blast Radius | Small. Affects a limited scope — one service, one team, one page. |
| Information Asymmetry | High. Real-world data beats pre-decision analysis. |
| Time Sensitivity | Often high. Delaying costs more than making a suboptimal choice. |
Concrete Engineering Examples:
- Changing a button color or UI layout — deploy it, measure click-through rates, revert if metrics drop.
- Deploying a new feature behind a feature flag — the flag itself is the reversal mechanism. If the feature causes errors, disable it in seconds without a new deployment.
- Trying a new CLI tool or library on a non-critical path — if
ripgrepdoesn't work better thangrepfor your use case, stop using it. Nothing changed. - Adjusting resource limits on a stateless container — increase memory from 256Mi to 512Mi. If it doesn't help, change it back. Kubernetes applies the change on the next pod restart.
- Switching from JSON to YAML for a configuration file — a formatting preference with zero runtime impact that can be changed at any time.
- Adding a new monitoring alert threshold — if it's too noisy, adjust or delete it.
Strategy: Bias Toward Action
For Type 2 decisions, the optimal process is: make the call, ship it, observe the results, adjust. Do not require approval from a committee. Let individuals or small teams decide. The goal is to maximize the number of experiments per unit of time, because each experiment generates real data that no amount of pre-decision analysis can replicate.
When asked "How do you make decisions quickly without being reckless?", describe the Type 1/Type 2 framework explicitly. Explain that you identify the reversal cost first, then match the process to the decision type. Interviewers at companies like Amazon, Netflix, and Stripe specifically look for this kind of structured decision-making vocabulary. A strong answer: "I ask — what's the worst case, and can we undo it? If yes, I bias toward trying it. If no, I slow down and bring in more perspectives."
Key Point: The most common organizational failure is not making bad Type 2 decisions — it's making Type 2 decisions too slowly. Every hour spent in a meeting debating a reversible choice is an hour of experimentation lost. Senior engineers protect their team's velocity by actively identifying Type 2 decisions and short-circuiting unnecessary process around them.